Machine Learning Model Drift: Predicting Diagnostic Imaging Follow-Up as a Case Example

There is no author summary for this article yet. Authors can add summaries to their articles on ScienceOpen to make them more accessible to a non-specialist audience.

Abstract

Address model drift in a machine learning (ML) model for predicting diagnostic imaging follow-up using data augmentation with more recent data versus retraining new predictive models.

Related collections

Most cited references 44

Record: found
Abstract: found
Article: found

Is Open Access

Comparing different supervised machine learning algorithms for disease prediction

Shahadat Uddin, Arif Khan, Md Ekramul Hossain … (2019)

Background Supervised machine learning algorithms have been a dominant method in the data mining field. Disease prediction using health data has recently shown a potential application area for these methods. This study ai7ms to identify the key trends among different types of supervised machine learning algorithms, and their performance and usage for disease risk prediction. Methods In this study, extensive research efforts were made to identify those studies that applied more than one supervised machine learning algorithm on single disease prediction. Two databases (i.e., Scopus and PubMed) were searched for different types of search items. Thus, we selected 48 articles in total for the comparison among variants supervised machine learning algorithms for disease prediction. Results We found that the Support Vector Machine (SVM) algorithm is applied most frequently (in 29 studies) followed by the Naïve Bayes algorithm (in 23 studies). However, the Random Forest (RF) algorithm showed superior accuracy comparatively. Of the 17 studies where it was applied, RF showed the highest accuracy in 9 of them, i.e., 53%. This was followed by SVM which topped in 41% of the studies it was considered. Conclusion This study provides a wide overview of the relative performance of different variants of supervised machine learning algorithms for disease prediction. This important information of relative performance can be used to aid researchers in the selection of an appropriate supervised machine learning algorithm for their studies.

0 comments Cited 253 times – based on 0 reviews      Review now

Bookmark

Record: found
Abstract: not found
Article: not found

Scikit-learn: machine learning in python

F Pedegrosa (2011)

0 comments Cited 237 times – based on 0 reviews      Review now

Bookmark

Record: found
Abstract: found
Article: not found

Derivation of a simple clinical model to categorize patients probability of pulmonary embolism: increasing the models utility with the SimpliRED D-dimer.

P. S. Wells, D Anderson, M. Rodger … (2000)

We have previously demonstrated that a clinical model can be safely used in a management strategy in patients with suspected pulmonary embolism (PE). We sought to simplify the clinical model and determine a scoring system, that when combined with D-dimer results, would safely exclude PE without the need for other tests, in a large proportion of patients. We used a randomly selected sample of 80% of the patients that participated in a prospective cohort study of patients with suspected PE to perform a logistic regression analysis on 40 clinical variables to create a simple clinical prediction rule. Cut points on the new rule were determined to create two scoring systems. In the first scoring system patients were classified as having low, moderate and high probability of PE with the proportions being similar to those determined in our original study. The second system was designed to create two categories, PE likely and unlikely. The goal in the latter was that PE unlikely patients with a negative D-dimer result would have PE in less than 2% of cases. The proportion of patients with PE in each category was determined overall and according to a positive or negative SimpliRED D-dimer result. After these determinations we applied the models to the remaining 20% of patients as a validation of the results. The following seven variables and assigned scores (in brackets) were included in the clinical prediction rule: Clinical symptoms of DVT (3.0), no alternative diagnosis (3.0), heart rate >100 (1.5), immobilization or surgery in the previous four weeks (1.5), previous DVT/PE (1.5), hemoptysis (1.0) and malignancy (1.0). Patients were considered low probability if the score was 4.0. 7.8% of patients with scores of less than or equal to 4 had PE but if the D-dimer was negative in these patients the rate of PE was only 2.2% (95% CI = 1.0% to 4.0%) in the derivation set and 1.7% in the validation set. Importantly this combination occurred in 46% of our study patients. A score of <2.0 and a negative D-dimer results in a PE rate of 1.5% (95% CI = 0.4% to 3.7%) in the derivation set and 2.7% (95% CI = 0.3% to 9.0%) in the validation set and only occurred in 29% of patients. The combination of a score < or =4.0 by our simple clinical prediction rule and a negative SimpliRED D-Dimer result may safely exclude PE in a large proportion of patients with suspected PE.